Available online at www.sciencedirect.com Procedia Engineering 34 (2012 ) 640 645 9 th Conference of the International Sports Engineering Association (ISEA) Fatigue design of welded bicycle frames using a multiaxial criterion Alexandre Callens a, André Bignonnet b a Decathlon, MKniX Engineering Center, 59000 Lille, France b André Bignonnet Consulting, 49750 Beaulieu, France Accepted 02 March 2012 Abstract This paper describes the methodology developed by Decathlon, through its MKniX Engineering Center, to master the fatigue design of welded aluminium-alloy bicycle frame. The objective is to optimize the design prior to the standard testing by calculating the fatigue reliability of the bicycle frame. The fatigue assessment method is based on the Dang Van multiaxial fatigue criterion combined with a unique S-N design curve independent of the geometry of the welded structure and the loading mode. The design stress is defined through a linear elastic finite element analysis using a specific thin shell meshing method. 2012 Published by Elsevier Ltd. Open access under CC BY-NC-ND license. Keywords: Bicycle frame; fatigue; multiaxial; weld; thin shell meshing 1. Introduction Bicycle frames are covered by a normative context (European standard EN14764, EN14766, and EN14781) which includes standardized fatigue testing. For years, the frame fatigue design has been assessed experimentally. Numerous physical prototypes had to be tested to optimize the product. This iterative experimental method was time consuming and costly. Classical finite element analysis has then been introduced. It was accurate to localize the critical area but not to predict the fatigue strength of the structure. Therefore a numerical fatigue and reliability assessment method was implemented to optimize the frame design, to reduce the time to market and the design cost. The objective is that each frame will pass the standard tests at the first time. In addition, to guarantee the high level quality of our products, the fatigue reliability has to be mastered precisely by an appropriate statistical process control. During very severe fatigue tests, fatigue failures always occur at the weld toes and classical weld fatigue analysis does not allow a correct prediction of the fatigue strength of the structure. We thus introduce the Dang Van criterion and a specific meshing method to assess the fatigue behaviour of our bicycle frames. 1877-7058 2012 Published by Elsevier Ltd. doi:10.1016/j.proeng.2012.04.109 Open access under CC BY-NC-ND license.
Alexandre Callens and André Bignonnet / Procedia Engineering 34 ( 2012 ) 640 645 641 2. Stress calculation by the structural approach A typical aluminium welded frame is made of thin wall tubes. The geometries of welded areas are rather complex, especially in the bottom bracket area. Moreover, bicycle frame welds are mainly handmade and there is a variety of weld size. These facts limit the application of classical weld fatigue analysis as presented in literature. The nominal stress approach, as for example in Eurocode 9 [1], classifies standard welded joint with a fatigue curve for each joint. Since welded joints may be different on each frame, this approach is not applicable. The geometric stress approach, as presented in API rules [2] or in IIW [3] is not well suited for thin shell structure. Indeed, aluminium tubes used for frames have a thickness between 1.4mm and 1.8mm with small cross section. For the same reason, the local approach presented in IIW [3] is also not suited as it would lead to a very small and complex meshing for an everyday industrial application. As welded areas are complex, with a variety of weld size and multiaxial stress state in critical areas, the method developed by J-L. Fayard [4, 5] has been chosen. This structural approach has the advantage to use a unique S-N design curve, independent of the geometry of the welded structure and of the loading mode. This method is also used in various industrial applications [6, 7]. Moreover, it particularly suits our context, with the advantage to take in account the weld size thanks to a specific meshing rule. The aim is to find a solution to determine the damage in areas where cracks initiate and lead to failures. In welded structures, when the industrial welding process is mastered to avoid a root failure, failures always occur at the weld toes. In these areas, local strains are not small and the material is not perfectly homogenous. The mechanical behaviour in this zone is therefore not accessible by the usual mechanics of solids. Nevertheless, it is possible to estimate an asymptotic solution at the periphery of this critical zone (figure 1). It is defined by a linear homogenous and isotropic elasticity in small strains calculation. This asymptotic value at the border of the critical zone allows the control and correct interpretation of the fatigue damage in the critical zone (see [4]). The calculation result will significantly depend on the modelling of the weld, because the asymptotic stress is in highly stressed zones with a sharp gradient. We thus apply a specific meshing methodology that takes into account the weld influence (local stiffness and size). The asymptotic design stress may be determined from a finite element calculation for every welded joint. As it uses the stress and strain analysis from a simple structural calculation, the above described method is more accessible for an everyday engineering application than fracture mechanics or local approach. Fig. 1. Asymptotic solution
642 Alexandre Callens and André Bignonnet / Procedia Engineering 34 ( 2012 ) 640 645 3. Computing procedure The computing procedure is based on a midsurface thin shell modelling. The stress flow is going from one sheet to the other through the weld. This is represented by rigid body elements which links the two shells by the middle of the weld side (figure 2). Fig. 2. Modelling method The design stress is calculated at the centre of gravity of the element at top face without node interpolation. All elements have the size of the fusion penetration length. Consequently, the meshing and so the design stress, directly depend on the shell thickness and the fusion penetration length. Strain gauge chains were used to measure the real stress gradient near the welding toe in critical areas. These measurements were compared to a 3D modelling and to the specific shell modelling (figure 3a). As a result, the methodology was validated as the specific meshing method fits to our framework and gives accurate results. Strain (um/um) 1400 1200 1000 800 600 Strain Gauge Chain Shell modelling with specific rigid link 3D modelling 400 200 0 2 4 6 8 10 12 14 Lenght from the weld toe (mm) Fig. 3. (a)comparison of the stress gradient between modelling and measurement, (b) Meshing application on a bottom bracket A new modelling methodology has been defined for our frame. The designer can create the model following the welding specification. For the meshing, the automatic Ansys Workbench meshing tool is used with some size adaptation to fit the geometry. An in-house routine automatically places the rigid elements between the welding lines. Figure 3b shows a meshing application on the bracket area.
Alexandre Callens and André Bignonnet / Procedia Engineering 34 ( 2012 ) 640 645 643 4. Fatigue assessment 4.1. Multiaxial fatigue criterion Finite element analysis shows that actual loading and standard tests lead to multiaxial stress at critical points in welded areas. This guides us to the application of a multiaxial fatigue criterion. The Dang Van criterion is chosen due to its efficiency in multiaxial fatigue context. Moreover its characterization requires only two parameters, accessible with simple fatigue tests. Furthermore, it may be implemented in any finite element software, so the implementation perfectly fits our finite element analysis software Ansys Workbench. Thus it is usable by designers who are not fatigue experts. This fatigue criterion described in [5] is based on the ability to represent microscopic phenomenon, that pilot fatigue failure, by using macroscopic values calculated in design office through finite element calculation. It is defined by a p diagram (, shear amplitude and p, Hydrostatic pressure). The fatigue strength is represented in this diagram with a straight line defined by:. p B (1) where and B are material parameters. Those parameters are defined from fatigue tests results (described below in 4.2) with two loading modes, pure bending and pure torsion for example. The cyclic stress path in critical areas of the structure is determined by calculating (t) and p (t) at each loading step. (cycle) is the local shear amplitude for a given material plane and it requires the determination of the plane on which the set ( (cycle), p (t)) is a maximum for a load fatigue cycle. This is done by identifying on each shear plane, the smallest circle circumscribe to the shear path. The stress path is compared to the fatigue design line to determine the capacity of the structure to be safe or not. Fig. 4. Dang Van diagram The fatigue criterion can be written as following: max ( cycle). p( t) (2) t B For practical purposes, a danger coefficient, C D is introduced, C D =1/B. ( c - p-b) where B is the origin ordinate of the Design line, c is the projection of the maximum of the stress path for a cycle. Then, if C D >0, the failure risk is higher than the target reliability, the design is rejected. If C D =0, the design is optimized with regards to the target reliability. Finally, If C D <0, the failure risk is lower than the target reliability, the design is accepted, but it can be still optimized. The next paragraph gives an illustration of the C D colour map of the fatigue analysis results and the Dang Van diagram for the worst elements on each welding line (see 4.3 Application to an aluminium bicycle frame).
644 Alexandre Callens and André Bignonnet / Procedia Engineering 34 ( 2012 ) 640 645 4.2. The unique S-N curve The unique S-N curve was defined using fatigue tests on elementary welded tubular T joints. The specimens were welded in a real production way to be representative of the bicycle frame. Fatigue tests with two loading modes (pure bending, pure torsion) have been done to establish the Dang Van criterion for different number of cycles to failure (figure 5b). From these curves an S-N representation (figure 5a) is possible, taking the Design Stress S as c - p. 800 700 600 500 400 S= - p 50% curve Design curve Test data Bending Torsion Fatigue Criterion 300 200 (MPa) 100 0 N (cycle) 1,00E+04 1,00E+05 1,00E+06 1,00E+07 P (MPa) P Fig. 5. (a) SN curve defined by tests, (b) Dang Van diagram for bending and torsion test The unique S-N curve defined through elementary welded tubular joint has also been confirmed by a posteriori analysis that integrates multiple old frame fatigue results. This allows to reinforce our results and to confirm that a unique elementary welded tubular joint is sufficient and correct. Frames normative fatigue tests have to pass 50.000 cycles or 100.000 cycles depending on the standard. As a result, we have defined two Dang Van fatigue Design lines, one for 50.000 cycles and one for 100.000 cycles. Those Design lines are defined for a given level of reliability. They can be adapted, depending on the supplier, on the quality policy and so on. 4.3. Application to an aluminium bicycle frame To test the numerical fatigue assessment approach, the method was applied through a simulation of a pedalling standard fatigue test. The Dang Van danger" coefficient map is shown on figure 6. The elements appear in red if the failure risk is higher than what is required for the target of reliability. -0,0589 Max Stress Path Fatigue Criterion Fig. 6. Results on a pedalling fatigue test Bottom bracket area P
Alexandre Callens and André Bignonnet / Procedia Engineering 34 ( 2012 ) 640 645 645 On this example, there is no dangerous zone as all elements are below the Fatigue Design line. Nevertheless, the calculation shows us where the maximum stressed areas are and gives us the associated fatigue diagram. In reality this frame passes the standard test as calculated, however it has been decided to push some tests until failure. Fig. 7. Fatigue crack location with pedalling test continued to failure The failure mode during the pedalling test is at the area predicted by calculation. This has been done for numerous different new bicycle frames, and also on old frames that did not pass the new standard tests. The results were always in accordance with reality, giving the right failure area and level of reliability. 5. Conclusion The methodology presented above has been validated on bicycle frames and the fatigue strength prediction is excellent with regards to the standard tests. The use of the Dang Van fatigue criterion is well adapted to the multiaxial stress in critical areas of bicycle frames, and is easily characterized with simple fatigue tests. The automatic meshing developed was proven to be accurate for an industrial application. As it is dependent on fusion penetration length, it gives a powerful tool to link design and process. The methodology is now included in the reliability assessment procedure at Decathlon. Design and validation of frames are now realized in one-shot. This has a huge influence on the product quality, on the time to market and on the design cost. References [1] Eurocode 9, NF EN 1999-1-3, septembre 2007. [2] API Recommended Practice 2A WSD 21st Edition, Oct 2005. [3] IIW Document IIW-1823-07, December 2008, Recommendations for Fatigue Design of Welded Joints and Components. [4] J.L. Fayard, «Dimensionnement à la fatigue polycyclique de structures soudées», Thèse doct. Ecole Polytechnique, 1996. [5] Dang Van, K., Bignonnet, A.,Fayard,J-L., Assessment of Welded Structures by a Structural Multiaxial Fatigue Approach, In Biaxial/Multiaxial Fatigue and Fracture, Ed.A Carpinteri, M. De Freitas, A. Spagnoli, ESIS/STP31, pp3-22, Elvesier 2003. [6] A.Bignonnet, K.Dang Van, Application of the structural stress method for fatigue design of welded structures. Fatigue Design 2007, 21-22 Nov 2007, Senlis. [7] F. Conti, L. Verney, A. Bignonnet, Fatigue assessement of tubular welded connections with the structural stress approach. Fatigue Design 2009, 25-26 Nov 2009, Senlis.